| |
By Laz Foley, Fluent Inc. and Mel Cahoon, Volvo Penta of the Americas, Chesapeake, VA
View the pdf of this article
Today, CFD is extensively used to solve a variety of complex,
and sometimes computationally expensive transient flow
problems. While computer technology has advanced significantly,
CFD algorithms continue to be developed and
refined to ensure that transient efficiency is improved.
To date, the segregated solver in all versions of FLUENT has
made use of an iterative time advancement (ITA) scheme, where
a global iteration on the momentum and pressure (and other)
equations is performed many times within each time step. The
release of FLUENT 6.2 will provide two new second-order, noniterative
time advancement (NITA) schemes: the non-iterative
PISO [1] method and the non-iterative fractional step [2]
method. These NITA algorithms obviate the need for multiple,
global iterations within each time-step. Instead, a number of subiterations
are performed on the individual equations to ensure
second-order spatial accuracy. Because global iterations are more
time-consuming (and therefore expensive) than the sub-iterations,
the overall computational cost of the new NITA algorithms
is significantly less than the fully iterative time advancement algorithm.
While the NITA schemes run several times faster than the
iterative schemes, the non-iterative fractional step algorithm,
which uses the SIMPLEC pressure-velocity coupling corrector,
can offer up to an additional 20% improvement in transient efficiency
compared to the non-iterative PISO algorithm.
The intake manifold
geometry
Static pressure contour comparisons between FLUENT 6.1 (left, using the
iterative algorithm) and FLUENT 6.2 (right, using the non-iterative PISO
algorithm) in a V8 intake manifold
As an example of this new capability, a V8 intake manifold,
typical of that used in NASCAR racing, has been simulated. The
manifold has one inlet, where the carburetor is mounted, and
eight outlets, which are intake runners for the engine cylinders.
The mesh contains 155,000 cells, and the boundary conditions
– for the inlet and all eight outlets – are prescribed using a transient
table. The transient table data, which consists of mass flux
and temperature values, was acquired from a 1D engine cycle
simulation.
Average static pressure versus crank angle at the intake manifold inlet
boundary, showing good agreement between the ITA and NITA-PISO
algorithms
The FLUENT results have been used to provide valuable
information about the nature of the manifold flow field, including
total pressure losses and flow splits for the manifold runners.
They have also been used to draw comparisons between
the predictions of the iterative and non-iterative solvers. Static
pressure contours illustrate that there are minor differences
(~20 Pa) between the flow fields predicted by the iterative
(FLUENT 6.1) and non-iterative PISO (FLUENT 6.2) schemes,
where the latter was four times faster to solve than the former.
Results such as these illustrate that the staggering improvement
in transient efficiency can be achieved with insignificant compromise
in numerical accuracy.
References:
- R.I. Issa, “Solution of the Implicitly Discretized Fluid Flow Equations by
Operator-Splitting,” Journal of Computational Physics 62,
p. 40-65, 1985.
- J.B. Perot, “An Analysis of the Fractional-Step Method,” Journal of
Computational Physics 108, p. 51-58, 1993.
|
|
|
